Gcforest-Based Fault Diagnosis Method For Rolling Bearing

At present, the intelligent diagnosis technology of bearing is mostly dependent on the deep neural network, which has high requirements for computer hardware, especially CPU. It also requires a large enough training data set with labels and a large number of hyperparameters. To solve these problems, a rolling bearing intelligent diagnosis method based on gcforest (multi-Grained Cascade forest) theory is proposed in this paper. First, the data signals collected by the experiment are preprocessed, and stored in the data set in the form of numerical values; then the data in the data set is directly input into the model for training or recognition. Experiments show that this method has good fault recognition accuracy, and it can also achieve satisfactory results by training with small-scale data sets. At the same time, it has few hyperparameters, and the training process has very low requirements for the hardware.